Re: [PATCH for-rc] RDMA/core: Call dev_put after query_port in iw_query_port

[Jason Gunthorpe <jgg@ziepe.ca>]

On Fri, May 19, 2023 at 11:11:19AM +0800, Guoqing Jiang wrote:
> There is a UAF report by syzbot.
> 
> BUG: KASAN: slab-use-after-free in siw_query_port+0x37b/0x3e0 drivers/infiniband/sw/siw/siw_verbs.c:177
> Read of size 4 at addr ffff888034efa0e8 by task kworker/1:4/24211
> 
> CPU: 1 PID: 24211 Comm: kworker/1:4 Not tainted 6.4.0-rc1-syzkaller-00012-g16a8829130ca #0
> Hardware name: Google Compute Engine/Google Compute Engine, BIOS Google 04/14/2023
> Workqueue: infiniband ib_cache_event_task
> Call Trace:
>  <TASK>
>  __dump_stack lib/dump_stack.c:88 [inline]
>  dump_stack_lvl+0xd9/0x150 lib/dump_stack.c:106
>  print_address_description.constprop.0+0x2c/0x3c0 mm/kasan/report.c:351
>  print_report mm/kasan/report.c:462 [inline]
>  kasan_report+0x11c/0x130 mm/kasan/report.c:572
>  siw_query_port+0x37b/0x3e0 drivers/infiniband/sw/siw/siw_verbs.c:177
>  iw_query_port drivers/infiniband/core/device.c:2049 [inline]
>  ib_query_port drivers/infiniband/core/device.c:2090 [inline]
>  ib_query_port+0x3c4/0x8f0 drivers/infiniband/core/device.c:2082
>  ib_cache_update.part.0+0xcf/0x920 drivers/infiniband/core/cache.c:1487
>  ib_cache_update drivers/infiniband/core/cache.c:1561 [inline]
>  ib_cache_event_task+0x1b1/0x270 drivers/infiniband/core/cache.c:1561
>  process_one_work+0x99a/0x15e0 kernel/workqueue.c:2405
>  worker_thread+0x67d/0x10c0 kernel/workqueue.c:2552
>  kthread+0x344/0x440 kernel/kthread.c:379
>  ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:308
>  </TASK>
> 
> It happened because netdev could be freed if the last reference
> is released, but drivers still dereference netdev in query_port.
> So let's guard query_port with dev_hold and dev_put.
> 
> Reported-by: syzbot+79f283f1f4ccc6e8b624@syzkaller.appspotmail.com
> Closes: https://lore.kernel.org/lkml/0000000000001f992805fb79ce97@google.com/
> Signed-off-by: Guoqing Jiang <guoqing.jiang@linux.dev>
> ---
> I guess another option could be call ib_device_get_netdev to get
> netdev in siw_query_port instead of dereference netdev directly. 
> If so, then other drivers (irdma_query_port and ocrdma_query_port)
> may need to make relevant change as well.

Something is wrong in siw if it is UAF'ing it's own memory:

	attr->max_mtu = ib_mtu_int_to_enum(sdev->netdev->mtu);

It needs to protect sedv->netdev somehow on its own.

Jason